Re: [PATCH v4 7/9] x86/dmop: Add XEN_DMOP_{bind,unbind}_pt_msi_irq DM ops

[Jan Beulich <jbeulich@suse.com>]

On 28.04.2026 18:02, Teddy Astie wrote:
> Le 27/04/2026 à 15:57, Julian Vetter a écrit :
>> Add two DM ops for MSI passthrough IRQs. These new DM ops take the raw
>> MSI address and data fields rather than pre-decoded gflags values. Xen
>> decodes the destination ID via msi_addr_to_gflags(), including any
>> extended destination bits in address[11:5]. This means the device model
>> does not need to understand the extended destination ID encoding, and
>> simply forwards the MSI address it observes from the guest.
>>
>> With these DM ops in place, redirect xc_domain_update_msi_irq() and
>> xc_domain_unbind_msi_irq() in libxenctrl to use
>> xendevicemodel_bind_pt_msi_irq() / xendevicemodel_unbind_pt_msi_irq()
>> via xch->dmod. The gflags/gvec arguments are translated to the raw MSI
>> address and data words at the libxc level using the standard x86 MSI
>> address format.
>>
>> Reject the PT_IRQ_TYPE_MSI sub-case in XEN_DOMCTL_bind_pt_irq and
>> XEN_DOMCTL_unbind_pt_irq: all callers now go through the DM op path, so
>> the domctl sub-case is fully obsolete.
> 
> We probably want to reflect that on XEN_DOMCTL_{un}bind_pt_irq interface 
> in domctl.h (e.g through a note saying that PT_IRQ_TYPE_MSI type is now 
> deprecated and unsupported).

Which may further want mentioning in ./CHANGELOG.md.

>> --- a/tools/libs/devicemodel/core.c
>> +++ b/tools/libs/devicemodel/core.c
>> @@ -645,6 +645,44 @@ int xendevicemodel_nr_vcpus(
>>       return 0;
>>   }
>>   
>> +int xendevicemodel_bind_pt_msi_irq(
>> +    xendevicemodel_handle *dmod, domid_t domid, uint32_t machine_irq,
>> +    uint64_t msi_addr, uint32_t msi_data, uint64_t gtable, int unmasked)
>> +{
>> +    struct xen_dm_op op;
>> +    struct xen_dm_op_bind_pt_msi_irq *data;
>> +
>> +    memset(&op, 0, sizeof(op));
>> +
>> +    op.op = XEN_DMOP_bind_pt_msi_irq;
>> +    data = &op.u.bind_pt_msi_irq;
>> +
>> +    data->machine_irq = machine_irq;
>> +    data->data = msi_data;
>> +    data->addr = msi_addr;
>> +    data->gtable = gtable;
>> +    if ( unmasked )
>> +        data->flags |= XEN_DMOP_MSI_FLAG_UNMASKED;
>> +
>> +    return xendevicemodel_op(dmod, domid, 1, &op, sizeof(op));
>> +}
>> +
>> +int xendevicemodel_unbind_pt_msi_irq(
>> +    xendevicemodel_handle *dmod, domid_t domid, uint32_t machine_irq)
>> +{
>> +    struct xen_dm_op op;
>> +    struct xen_dm_op_unbind_pt_msi_irq *data;
>> +
>> +    memset(&op, 0, sizeof(op));
>> +
>> +    op.op = XEN_DMOP_unbind_pt_msi_irq;
>> +    data = &op.u.unbind_pt_msi_irq;
>> +
>> +    data->machine_irq = machine_irq;
>> +
>> +    return xendevicemodel_op(dmod, domid, 1, &op, sizeof(op));
>> +}
>> +
> 
> I think we want to mark 
> xc_domain_update_msi_irq/xc_domain_unbind_msi_irq as deprecated since we 
> implemented a newer (better) version of it in xendevicemodel; and the 
> old one is now a wrapper.

Why mark it deprecated? It can be removed right away when there are no callers
left. libxc doesn't offer a stable API.

Jan